[hts-users:02669] [Algorithm] The context cluster decision tree for adaptation
Hi, sorry to bother you for such a algorithmic question
QS1. In the resulting dec.tree and dec.class, I got a very big number of classes, which is only a little less than the training decision tree. Shouldn't it be much less than the decision tree because of the lack for adaptation data?
QS2. In the resulting xform file, the number of xforms magnificantly decreased from the number of dec.tree class. Shouldn't every base class be assigned with a transformation?
Thanks in advance.
--
Na Xingyu
Beijing Institute of Technology